R devel - Jul 2011 - R CMD CHECK doens't run configure when testing install? (Revised)

I'm trying to get ready to submit a package to CRAN, but in order for the
package to install on OS X, I need to temporarily set the environment variable
NOAWT=1.  I put 'export NOAWT=1' in my package's 'configure'
script, and 'R CMD INSTALL MyPackage' returns with no errors.   However
when I do 'R CMD CHECK MyPackage', the installation test fails, because
the configure script is not invoked.  If I set the variable manually in Terminal
before I run CHECK, then it passes all the tests.

Is the configure script the right place to be setting this variable?

If so, why isn't configure being invoked as part of R CMD CHECK?

If not, where should I set the environmental variable NOAWT=1 such that the
change will be reflected in the R CMD CHECK installation test?

Here is the source tar-ball of my package if you want to try for yourself.
http://www.cs.ucla.edu/~alexalex/R/src/contrib/DeducerText_0.1-0.tar.gz

Thanks.

--
Alex Rickett
ack.vandal at gmail.com

Please don't spam multiple lists!

And follow the posting guide: you have not given the reproducible 
example asked for, there is no signature block ....

Of course you could waste people's time by submitting a package that 
does not past R CMD check, but you are asked to follow a set of checks 
in 'Writing R Extensions' and the CRAN workers do not have unlimited 
patience ....

On Thu, 28 Jul 2011, Alexander James Rickett wrote:
> I'm trying to get ready to submit a package to CRAN, but in order 
> for the package to install on OS X, I need to temporarily set an 
> environment variable.  I put this in the 'configure' script, and
'R
> CMD INSTALL MyPackage' works fine, but when I do 'R CMD CHECK 
> MyPackage', and it tests installation, the configure script doesn't
> run and consequently the installation fails.  Should I be setting 
> the variable another way?  It passes all the other checks, and it 
> will install outside of check, so could I just submit it as is?
>
> Thanks!
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

On Jul 28, 2011, at 6:56 PM, Alexander James Rickett wrote:
> I'm trying to get ready to submit a package to CRAN, but in order for
the package to install on OS X, I need to temporarily set the environment
variable NOAWT=1.  I put 'export NOAWT=1' in my package's
'configure' script, and 'R CMD INSTALL MyPackage' returns with
no errors.   However when I do 'R CMD CHECK MyPackage', the installation
test fails, because the configure script is not invoked.  If I set the variable
manually in Terminal before I run CHECK, then it passes all the tests.
> 
> Is the configure script the right place to be setting this variable?
> 
> If so, why isn't configure being invoked as part of R CMD CHECK?
> 
> If not, where should I set the environmental variable NOAWT=1 such that the
change will be reflected in the R CMD CHECK installation test?
> 
You should not set it in the package at all. It is a run-time variable specific
to OS X checking so it is expected to be set in the OS and the CRAN checks on OS
X set it automatically.

If you are testing your package locally, you can simply use

NOAWT=1 R CMD check DeducerText_0.1-0.tar.gz

Cheers,
Simon

> Here is the source tar-ball of my package if you want to try for yourself.
> http://www.cs.ucla.edu/~alexalex/R/src/contrib/DeducerText_0.1-0.tar.gz
> 
> Thanks.
> 
> --
> Alex Rickett
> ack.vandal at gmail.com
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
> 
>